As part of the ongoing Open Correspondence rewrite, I’ve started working on some visualisations after a conversation with Rufus Pollock during one of the Humanities calls. One of the immediate ones was a force-directed graph to link all the correspondents to the authors. Well author at the moment. Although I am aware of SigmaJS, I decided to use D3‘s force directed graphs as an initial exploration in the use of such graphs in determining their use in the project.
It lays the basis for future graphs though.
I have put together two network graphs showing the total overview of the letters and one which looks at letters linked to the novel, Bleak House.
The first graph just shows the counts of authors linked to Charles Dickens. This has been set up limiting the group to two groups, the first being the root (in this case the author, Charles Dickens) and the second being an arbitrary number set for each link. Whilst it looks fairly uniform and boring, it does give a sense of the relationships as expressed in the number of letters to each recipient.
In the second graph, I changed through groups to being the known years in the letters, to limit it to a range between 1840 and 1872. I’m guessing that months could be added but it makes the potential group range rather large. Initial attempts just used the force arguments for charge and link. Normally, with a few groups this was fine but with a large amount of links and groups the graph became somewhat large and unreadable. To solve this, I read the D3 Force Gravity wiki and added gravity to bind the graph more closely together.
This, I think, gives the viewer a sense of the relationships.
An idea that I had was to change the groups so that the focus, or the target, was the year or decade but this was very confusing and does not give a sense of cohesion or collection. Although it does give a quick overview of the groups of letters by decade, it distracts from the original set of relationships which might give a user a better understanding of them. I suspect that time dimensions are better and more clearly expressed using time lines and charts but this is a set of experiments.
Both give a good overview of the figures but what happens if we are looking for a particular novel. One of the aims of the Open Correspondence project was to explore the ways in which information about texts, such as novels or plays, was transmitted. Effectively looking for hash tags in Twitter where these are used to highlight a word, set of words or phrases.
I went back to the original dataset and re-parsed them looking for all mentions of Bleak House. There were not as many as I had thought but the collection that I have only contains around 900 letters which is not the full dataset. However I expressed the relationships as a graph and a time line (not yet published) to show how we might want to visualise them. They do show an outlier, a letter to the author Wilkie Collins in 1862, but that the rest are in 1852 and 1853 when the novel was being serialised and then published in book form.
I did create a quick time chart but I have not yet published it because it is a little too small and messy but it will appear shortly. What it does show, unsurprisingly is the publications spark a cluster of letters. I also ran the numbers for the publication of David Copperfield and this showed a similar pattern but the clusters are longer and more sustained. At first glance, this might show that Copperfield was more popular, that the collection editors (Georgina Hogarth and Mamie Dickens) either kept or had more letters pertaining to this novel, or it might be some other reason.
The next post will explore time lines and charts since D3 can be used to create these as can Simile, a project that I have used before but not for some time.